Fuzzy contextual inference system for medical image enhancement

Abstract

PurposeIn order to facilitate early detection and further diagnosis, enhancing contrast and preserving natural characteristics are the fundamental requirement. In MR imaging, the poor quality image specifically the low contrast image may deliver inadequate data for the visual interpretation of affected portions. MethodsFuzzy Contextual Dissimilarity Adaptive Histogram Equalization (FCDAHE) is employed to enhance the contrast of medical images. The proposed method consists of two modules. In the first module, fuzzy dissimilarity histogram is formulated by finding fuzzy neighbourhood dissimilarity for every pixel using fuzzy rules. Then, the fuzzy dissimilarity clip limit is computed based on a fuzzy inference system. Contrast (C) and discrete entropy (E) are the two parameters used by the Fuzzy Inference System (FIS) to achieve fuzzy dissimilarity clip limit. Then, the enhanced output is obtained by equalizing the fuzzy dissimilarity adaptive histogram. In the second module, Contextual Intensity Transformation (CIT) is applied to fuzzy dissimilarity adaptive histogram equalized output to get final enhanced images. ResultsExperiments are conducted and tested on a wide variety of MR images extracted from BRATS 2015 database and Magnetic Resonance – Technology Information Portal to evaluate the performance of the proposed method both qualitatively and quantitatively. Extensive quantitative measures show that the proposed technique yields better performance in terms of PSNR, entropy, contrast ratio and EME when compared to state of art enhancement methods. ConclusionThe proposed FCDAHE algorithm not only improves contrast but also preserves natural characteristics. The proposed method offers a better scope for disease analysis and diagnosis.